Adaptation to vertical disparity induced

advertisement
Vision Research 43 (2003) 135–147
www.elsevier.com/locate/visres
Adaptation to vertical disparity induced-depth: implications
for disparity processing
Philip A. Duke *, Laurie M. Wilcox
Centre for Vision Research, York University, 4700 Keele Street, Toronto, Ontario, Canada M3J 1P3
Received 24 April 2002; received in revised form 31 July 2002
Abstract
Depth aftereffects produced by prolonged inspection of an object in depth can be mediated by monocular and binocular depth
cues. The adapting mechanisms responsible for such effects have not yet been fully determined. Theories of binocular depth
aftereffects typically posit a role of an adaptive horizontal disparity sensitive mechanism, implying multiple cue-specific mechanisms
for depth aftereffects. Here we examined whether binocular depth aftereffects can be attributed to such a cue-specific mechanism. In
Experiment 1 we did so using a technique allowing us to maintain horizontal disparities and vergence constant for our adaptation
stimuli, whilst manipulating simulated depth by virtue of a vertical disparity induced-depth effect. We found that depth aftereffects
were almost identical to those produced by adaptation to stimuli of equivalent depth produced by conventional horizontal disparity
modulations. In Experiment 2, we examined depth aftereffects following adaptation to apparently frontal surfaces produced by
different combinations of horizontal and vertical disparity modulations. Aftereffects were close to zero. These results suggest that
binocular depth aftereffects are not due to adaptation of a horizontal disparity sensitive mechanism, and we argue that adaptation
occurs at the level of a 3D shape sensitive mechanism derived from multiple cues. Experiment 3 was a control to examine whether
the two types of adaptation stimuli in Experiment 1 were indeed perceptually the same, since in theory they may differ if vertical
disparities influenced metric depth scaling. We found no evidence of this, and concluded that the two classes of stimuli used in
Experiment 1, though consisting of very different patterns of disparity, were perceptually equivalent.
Ó 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Stereopsis; Vertical disparity; Horizontal disparity; Induced-depth; Depth aftereffect
1. Introduction
Prolonged viewing of an object in depth can produce
distortion of the apparent depth of subsequently viewed
objects. Such depth aftereffects were first reported by
K€
ohler and Emery (1947) who found that following
binocular inspection of a line slanted in depth for a period
of 2 min, a frontal test line appeared slanted in the opposite direction from the inspection stimulus. Similarly,
they found that binocular inspection of a curved surface
in depth caused a frontoparallel test surface to appear
oppositely curved in depth. Whilst these experiments did
not determine the cue(s) responsible for the aftereffects, it
has since been shown that depth aftereffects can be mediated by horizontal disparity and by monocular cues.
*
Corresponding author.
E-mail address: pduke@yorku.ca (P.A. Duke).
Evidence that monocular cues are sufficient to produce depth aftereffects was provided by Bergman and
Gibson (1959). They measured aftereffects of the apparent slant of a physically frontoparallel plane following inspection of a randomly textured, slanted planar
surface under binocular and monocular viewing conditions. The fact that the two conditions produced similar
aftereffects demonstrated that depth aftereffects can be
mediated by texture cues in the absence of disparity
information.
A depth aftereffect which can be attributable only to
horizontal disparity information was demonstrated by
Blakemore and Julesz (1971), using random dot stereograms. They used an inspection stereogram of two adjacent squares, one in front and one behind the fixation
plane, and a test stereogram of two adjacent squares
both in the fixation plane. Following inspection, the test
squares appeared to be displaced in depth in the opposite direction from the inspection squares. Again using
0042-6989/03/$ - see front matter Ó 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 2 - 6 9 8 9 ( 0 2 ) 0 0 4 9 2 - 3
136
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
random dot stereograms, Long and Over (1973) examined depth aftereffects over a set of inspection stimulus
depths. Their results were consistent with those of
Blakemore and Julesz, and further showed that the
magnitude of the aftereffect varied with the horizontal
disparity of the inspection stimulus roughly as an antisymmetric function with maximum effects at around 8
arcmin. The finding of such horizontal disparity-specific
depth aftereffects prompted explanations of the effect in
terms of adaptation of horizontal disparity sensitive
mechanisms (e.g. Blakemore & Julesz, 1971; Howard &
Rogers, 1995; Long & Over, 1973; Mitchell & Baker,
1973). Accounts of this type are attractive since depth
aftereffect phenomena share several analogies with tilt
aftereffect phenomena, for which an explanation in
terms of neural adaptation is more widely accepted (e.g.
Bednar & Miikkulainen, 1997; Blankemore & Campbell,
1969; Coltheart, 1971; Gibson & Radner, 1937). Under
this scheme, depth aftereffects from horizontal disparity
and those from pictorial depth cues result, at least in
part, from adaptation of different mechanisms. However, some evidence suggests that a common mechanism
plays a role.
Balch, Milewski, and Yonas (1977) found that depth
aftereffects could result when inspection and test objects
were defined by different types of depth cue, suggesting
that adaptation can occur at a higher level of processing,
at or beyond the level of depth cue combination. They
examined transfer of a slant aftereffect between all four
combinations of monocular and binocular inspection
conditions, and monocular and binocular test conditions. Aftereffects (of different magnitudes) were obtained in all four conditions, suggesting that to some
degree, adaptation occurred in a common mechanism
which is independent of the type of depth cue specifying the surface slant. Poom and B€
orjesson (1999) also
examined transfer of depth aftereffects between cues,
and obtained similar results. They argued that the
differences in aftereffect magnitude found for different
combinations of adaptation and test stimulus depth cue
conditions were due to different adaptive changes in
the integration of the various depth cues, as opposed
to adaptation of low level (e.g. horizontal disparity) or
high level (e.g. 3D shape) representations. A recently
published study by Domini, Adams, and Banks (2001)
examined whether depth aftereffects for curved surfaces
could be attributed to adaptation a horizontal disparity
or a higher level 3D shape sensitive mechanism. They
did so using stimuli depicting a horizontal row of dots
lying on a curved surface, viewed at different distances
from the observer. One set of their stimuli produced the
same theoretical apparent shapes, from different patterns of horizontal disparity, and another produced
different theoretical apparent shapes from the same
pattern of horizontal disparity. Their experiment indicated a role of apparent shape in determining aftereffect
magnitude, and did not find support for an effect of
horizontal disparity adaptation in the conditions tested.
In summary, all of the mechanisms responsible for
depth aftereffects have not been identified with certainty.
Collectively, this evidence suggests a role of a 3D shape
sensitive mechanism which receives input from multiple
cues, and a possible role of mechanisms sensitive to
specific depth cues (e.g. horizontal disparity, texture).
In the present study we examined whether binocular
depth aftereffects can be attributed to adaptation of a
horizontal disparity sensitive mechanism, or to a higher
level 3D shape sensitive mechanism, using an apparent
frontoparallel plane (AFPP) task (e.g. Foley, 1980). In
Experiment 1 we examined aftereffects following adaptation to stimuli with the same horizontal disparity and
vergence, but different apparent depths, produced by a
vertical disparity induced-depth effect. We found that
aftereffects varied with apparent shape, so were not determined by the horizontal disparities of the stimulus. We
also compared these aftereffects with those produced by
adaptation to conventional horizontal disparity modulated stimuli of the same theoretical apparent depth as
the induced-depth stimuli. The data were approximately
identical, suggesting that apparent shape was the important factor in these depth aftereffects, and hence provides
support for a role of 3D shape sensitive mechanisms.
As a further test, in Experiment 2 we examined aftereffects following adaptation to a set of different disparity
patterns which all produced an apparently frontal stimulus. We found aftereffects close to zero, again suggesting
that apparent shape determines aftereffect magnitude.
In Experiment 3 we examined whether the conventional and induced-depth stimuli used in Experiment 1,
which were generated to give the same apparent depth in
theory, produced equivalent depth percepts in practice.
The similarity of depth percepts produced by our conventional and induced-depth stimuli was evaluated by
comparing the two in a matching task. We found that
observed matching data were very close to theoretical
values. This suggested that different combinations of
horizontal and vertical disparity modulations can lead
to the generation of the same percept, and indicated that
vertical disparities had no measurable effect on metric
scaling of depth. This validated our method in Experiment 1, and reinforced support for apparent 3D shape
as the important determiner of depth aftereffects in the
present study.
1.1. Vertical disparity induced-depth curvature
Here we describe relationships between horizontal
disparities, vertical disparities and viewing distance, 1
1
For ease of exposition we show common expressions which are
approximations of the true stereo geometry, though all calculations of
disparity for illustrations and stimuli used exact geometry.
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
and how by exploiting them, stereoscopic stimuli can be
created which have different apparent curvatures in
depth, but have the same horizontal disparities and
viewing distance. This analysis provides us with a tool
with which to investigate adaptation to apparent depth
independently of horizontal disparity and vergence.
Under symmetric vergence, the horizontal disparity,
H (in radians), of a point in the scene is given by the sum
of (1) a component due to horizontal retinal eccentricity
(2) a component due to depth (Mayhew & LonguetHiggins (1982)).
137
ð1Þ
vex. By making the assumption that vertical disparities
do not influence metric scaling, 2 we can calculate the
theoretical apparent surface depths, Z 0 , for any given
vertical disparity simulated distance, ds , using (5). In this
equation, horizontal disparities, Hxds corresponding to a
frontal surface at a simulated distance ds (consistent
with the vertical disparities in the stimulus) are subtracted from the measured horizontal disparities, 3 H ,
arising from viewing a frontal plane at the fixation
distance, d. The resultant horizontal disparities are expressed as depth values by rearranging (2). An illustration of the theoretical induced-depth curvature effect is
shown in Fig. 1.
I
Hz ¼ 2 Z
d
ð2Þ
Z0 ¼
H ¼ Hx þ Hz
ð3Þ
The magnitude of the induced-depth curvature created
by manipulating V in the manner described has often
been measured using a depth nulling (i.e. AFPP) task: an
observer sets a stimulus to appear frontal by choosing a
pattern of horizontal disparities from a set of patterns
which correspond to frontal surfaces at different fixation
distances (i.e. Hx in (1) calculated for a set of d). Studies
have typically found that the observed effect is around
60%–90% of the theoretical effect (Berends & Erkelens,
2001a).
I
H x ¼ x2
d
I is interocular distance, d is distance to the fixation
point and Z is the depth of the point from the plane of
fixation in centimetres; x is the horizontal cyclopean
eccentricity of the point in radians. Hence, for a frontal
surface in the plane of fixation, where by definition
Hz ¼ 0, H is given by the component due to horizontal
eccentricity component alone (1). It is therefore evident
that a frontal surface at one fixation distance will have
a different pattern of H at another, i.e. the pattern of
H from a frontal surface at a given fixation distance
is identical to that from a family of curved surfaces at
different fixation distances. This geometric fact must be
embodied in the computation of depth from disparity in
order that frontal surfaces do not appear systematically
curved with varying fixation distance. It is known that
the vertical component of binocular disparities is used
by the visual system as one source of information with
which to achieve this constancy (Berends & Erkelens,
2001a; Frisby et al., 1999; Helmholtz, 1910; Rogers &
Bradshaw, 1995).
Under symmetric vergence, the vertical disparity of a
point in a scene is given by the Mayhew and LonguetHiggins (1982) approximation:
I
V ¼ xy
d
ð4Þ
where x and y are horizontal and vertical cyclopean
eccentricities, respectively. Eq. (4) shows that a pattern
of V is uniquely specified by the fixation distance and as
such is independent of depth. The use of the vertical
disparity field in the frontal plane constancy process can
be demonstrated by changing vertical disparities in a
stimulus to simulate fixation at another distance, ds ,
whilst keeping horizontal disparities constant. When
vertical disparities correspond to ds < d, a physically
frontal surface appears to have concave curvature in
depth about the vertical meridian. When vertical disparities correspond to ds > d, the surface appears con-
d2
H Hxds
I
ð5Þ
2. Experiment 1
In Experiment 1 we compared depth aftereffects following adaptation to two sets of stimuli which produced
the same theoretical apparent curved surfaces in depth,
though derived from different patterns of disparity. A
set of vertical disparity induced-depth adaptation stimuli produced the perception of curved surfaces by
modulation of vertical disparities, holding horizontal
disparities constant. Another set produced, in theory,
the same apparent curved surfaces by modulation of
horizontal disparities whilst holding vertical disparities
constant. Are stereoscopic depth aftereffects due to
horizontal disparity adaptation? Any aftereffect from
the induced-depth stimuli cannot be due to horizontal
disparity adaptation. Are aftereffects due to apparent
2
This assumption is justified insofar as evidence suggests that the
effect of vertical disparity on metric scaling for small-field stimuli is
small (Rogers & Bradshaw, 1993) or possibly non-existent (Frisby
et al., 1999). More importantly, no such effect is evident in the data
presented here.
3
The theory that the component of measured H arising due to
depth is computed by compensation for viewing geometry is termed
Ôdisparity correctionÕ by G
arding, Porrill, Mayhew, and Frisby (1995).
This is a geometrically correct solution of the frontal plane constancy
problem, which is needed to formulate our stimuli. The present study is
not intended as an explicit test of any theory of stereopsis.
138
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
Fig. 1. Theoretical appearance of a physically frontal plane viewed at a vergence distance, d ¼ 45 cm, where vertical disparities are manipulated to
correspond to fixation at a simulated distance, ds . Values of ds , and hence Z 0 , chosen here for illustration also relate to our stimuli in Experiment 1: the
three inner functions correspond to our induced-depth adaptation stimuli.
3D shape? Dissimilar aftereffects across the two conditions would indicate that this cannot wholly be the case.
2.1. Method
2.1.1. Subjects
Four volunteers participated in Experiment 1; two
males and two females, aged between 24 and 27. Observer
PD was one of the authors. All observers had normal or
corrected visual acuity, with stereoacuity of 30 arcsec or
less, measured using the Titmus Randot Stereoteste.
2.1.2. Apparatus and stimuli
Stimuli were stereograms displayed as red–blue anaglyphs on a 21 in. Apple Trinitron monitor, at a viewing
distance of 45 cm. Both adaptation and test stimuli
comprised fields of Gaussian dot elements arranged as
nine regularly spaced, slightly jittered columns as shown
in Fig. 2. These appeared variously concave or convex
by virtue of their horizontal and vertical disparities. One
example is shown schematically in Fig. 3. We describe
the general method for generating these stimuli in Appendix A. The angular subtense of the stimulus was
approximately 43° horizontal by 32° vertical. The diameter of the dots was approximately 0.5° . A central
fixation cross (1.2°) was present throughout the experiment. A tightly fitting head and chin rest was used to
minimise movement of the observerÕs head. Responses
were made via computer keyboard. The experiment was
performed in a light-proofed room, in which no features
other than those of the stimuli were visible.
2.1.2.1. Adaptation stimuli. We used two classes of adaptation stimuli which we label Ôinduced-depth adaptation stimuliÕ and Ôconventional adaptation stimuliÕ. The
induced-depth adaptation stimuli had horizontal disparities which were always consistent with viewing a
Fig. 2. An example of a stimulus display, shown schematically. Actual
stimuli comprised red (left eye image), blue (right eye image) Gaussian
dots on a black background. Here open circles correspond to the left
eyeÕs image, and filled for the right image. This is a vertical disparity
induced-depth stimulus (4:6 cm condition). Vertical disparities are
consistent with a simulated viewing distance less than the veridical
distance, whilst horizontal disparities are consistent with a frontal
surface at the veridical distance. Such a stimulus appears to have
concave curvature in depth, i.e. columns further from the centre appear progressively closer to the observer as in Fig. 3.
frontal surface at 45 cm, whilst vertical disparities were
consistent with viewing a frontal surface at one of three
simulated vergence distances: 26.8, 45 and 140.6 cm.
These manipulations corresponded to theoretical induced-depths of )4.6, 0 and 4.6 cm, respectively, where
depth is measured from the frontal plane, at a point on
the edge of the display screen (22.8°, or 18.9 cm). The
shape of these surfaces is shown in Fig. 1 as a plan view.
In our pilot studies, we found the jittered columns
configuration produced the largest induced-depth effects
of our various candidates. This is presumably because
the columns provided a weaker texture cue to frontoparallel than others, such as a uniform field of dots (see
Banks & Backus, 1998).
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
Fig. 3. A schematic illustration of the apparent shape of an adapting
stimulus (4:6 cm condition shown), which was the same in both
conventional and induced-depth adaptation conditions. The outline of
the surface and the plane of the screen are added here for clarity.
Conventional adaptation stimuli had veridical
vertical disparities, i.e. consistent with fixation at the
veridical distance of 45 cm. Horizontal disparities were
modulated to produce the same apparent surfaces in
depth as those in the induced-depth conditions. This was
achieved using an appropriate modulation (as in Eq. (1))
which was obtained experimentally; we explain how in
Section 2.1.3.
2.1.2.2. Test stimuli. To measure depth aftereffects, we
used Ôconventional test stimuliÕ. This set of test images
had veridical vertical disparities, whilst horizontal disparity modulations (as in Eq. (1)) simulated object
depths ranging from )6.7 to 6.7 cm in steps of 3.5 mm.
In order to equate the apparent depth of the adaptation stimuli across induced-depth and conventional
depth conditions, a further set of test images were included. We call these Ôinduced-depth test stimuliÕ. These
test stimuli had the same set of horizontal disparity
modulations as the conventional test stimuli just described, and the same vertical disparity modulations as
the induced-depth adaptation stimuli, of which there
were three levels.
2.1.3. Procedure
The experiment took the form of a 2-factor fully repeated measures design. The two factors were: adaptation stimulus type (induced-depth or conventional) and
depth ()4.6, 0 and 4.6 cm), resulting in 6 adaptation
stimulus conditions, which were tested separately in six
experimental sessions of approximately 45 min. By way
of practice, observers performed one full session, chosen
at random, prior to the experiment proper. No observers
139
had any difficulty seeing the stimuli or performing the
task. Experimental sessions were performed on different
days and in a pseudorandom order.
Each session consisted of three parts, separated by 1
min rest periods. In each part the observer performed a
depth nulling (i.e. AFPP) task, i.e. they were asked to
identify the apparent sign of the depth in a test stimulus
(concave or convex) . Stimuli were displayed for 2 s. In
each part we used a single staircase procedure to estimate the horizontal disparity-vergence defined depth of
a stimulus which appeared frontoparallel. Observers
were instructed not to move their eyes from the central
cross during the experiment, but to scan the cross so as
to minimise contrast aftereffects. The details of each of
the three parts are as follows.
In part 1, observers were presented with conventional
test stimuli. Following 28 reversals, these data were used
to obtain the preadaptation measure of apparent
frontoparallel (ZAFPP part1 ).
In part 2, observers were presented with induceddepth test stimuli, all at one of the three vertical disparity levels. Following 28 reversals, these data were
used to obtain a corresponding measure of apparent
frontoparallel (ZAFPP part2 ), i.e. the induced-depth effect
(plus any bias).
In part 3, observers first inspected an adaptation
stimulus for 4 min. This was either an induced-depth
adaptation stimulus at the same vertical disparity level as
in part 2 (i.e simulated depth levels )4.6, 0 or 4.6 cm) or a
conventional adaptation stimulus with the same theoretical apparent depth (obtained as we describe in the
next section). After the initial adaptation period, conventional test stimuli were presented followed by adaptation top-ups of 30 sec. Following 24 reversals (lessened
to reduce fatigue effects), these data were used to obtain
the postadaptation measure of apparent frontoparallel
(ZAFPP part3 ). Depth aftereffects were calculated as the
difference between pre- and postadaptation measures of
apparent frontoparallel for conventional test stimuli
(ZAFPP part3 ZAFPP part1 ).
2.1.3.1. Equating the apparent depth of induced-depth and
conventional adaptation stimuli. The inclusion of part 2
(described above) allowed us to measure the amount of
depth (defined by horizontal disparity and vergence)
required to null a given induced-depth stimulus to appear frontal. This measure was required in the conventional adaptation stimulus conditions in order to
generate conventional adaptation stimuli with the same
apparent depth as the induced-depth adaptation stimuli.
Note that we often find that AFPP settings for conventional stimuli (part 1) are not veridical (Z ¼ 0) but
there is a bias, i.e. a physically frontal plane can appear
somewhat curved. For this reason the AFPP setting
from part 1 was subtracted from the AFPP setting from
part 2. By changing the sign of the resulting depth value
140
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
(i.e. converting from nulled depth), we obtain the depth
of a conventional stimulus which theoretically appears
the same as its corresponding induced-depth adaptation
stimulus. We call this theoretically equated conventional
depth Zc , obtained using Eq. (6).
Zc ¼ ðZAFPP part2 ZAFPP part1 Þ
ð6Þ
In each conventional adaptation condition, Zc was calculated from the results of parts 1 and 2; a conventional
adaptation stimulus with this simulated depth was then
generated for use in part 3.
Note that because of the offset from Z ¼ 0 in observersÕ AFPP judgements in part 1, the simulated depth
values for the conventional adaptation stimuli were not
exactly the same as their induced-depth equivalents, so
strictly, the depth levels in this condition are nominal,
and we use them as such for the purposes of plotting our
data. Later in Experiment 3 we examine the accuracy of
this indirect cross-condition depth-matching approach.
We find that the apparent depth of conventional depth
stimuli generated in this way closely match the apparent
depth of the corresponding induced-depth stimuli.
2.2. Results and discussion of Experiment 1
Here we show the group mean depth aftereffect (Fig.
4(a)) and the group mean AFPP setting for the induceddepth test stimuli presented in part 2 of this experiment
(Fig. 4(b)). The group mean preadaptation AFPP depth
settings using conventional test stimuli (part 1) are not
shown; they were always close to zero. This was true for
the individual observer data with the exception of CW
whose settings were consistently close to )2 cm.
Group mean postadaptation AFPP settings for
conventional test stimuli presented in part 3 differed
systematically from the corresponding preadaptation
settings made in response to the identical stimuli in part
1, in both conventional and induced-depth adaptation
conditions. This is evident from the aftereffect data
plotted in Fig. 4(a), which clearly shows that the induced-depth adaptation stimuli produced aftereffects
which were almost identical to those produced by conventional adaptation stimuli. Individual observer aftereffects are plotted in Fig. 5(e)–(h) and show remarkable
similarity across these conditions. A 2-factor repeated
measures ANOVA revealed a significant main effect of
adaptation stimulus depth (df ð2; 6Þ, F ¼ 7:62, p < 0:05)
and no effect of adaptation stimulus type as a main effect
(df ð1; 3Þ, F ¼ 1:54, n.s.), or in interaction with depth
(df ð2; 6Þ, F ¼ 0:34, n.s.).
In Fig. 4(b) the group mean AFPP settings for vertical disparity induced-depth test stimuli presented in
part 2 are shown; these were close to theoretical values.
The individual observer data from part 2 are shown in
Fig. 5(a)–(d). Settings for three observersÕ closely matched theoretical values, though again CW showed a constant error of approximately )2 cm. As a measure of the
strength of the induced-depth effect, we used the mean
slope of the observed data. For our observers these
values were between )0.67 and )0.88, and the group
mean was )0.77, which is consistent with previous
studies (Adams et al., 1996; Berends & Erkelens, 2001a;
Rogers & Bradshaw, 1995).
The main finding of Experiment 1 was that depth
aftereffects produced by adaptation to the induceddepth stimuli were the same as conventional depth
aftereffects. This result suggests that these depth after-
Fig. 4. Results of Experiment 1. Graph (a) shows the group mean depth aftereffects for both the conventional (filled circles) and induced-depth (open
circles) adaptation stimulus conditions. The two classes of adaptation stimuli were matched for apparent depth (see Section 2.1.3) and aftereffect data
for same apparent depth conditions are plotted at the same points along the abscissa. Values on the abscissa are the simulated depths of the induceddepth adaptation stimuli (which were slightly different from the simulated depths of the conventional stimuli––see Section 2.1.3). Error bars show one
standard error. Values on the abscissa are the magnitudes of the theoretical induced depths of the induced-depth stimuli. Whilst part 2 was identical
in both conventional and induced-depth adaptation stimulus conditions, the data from the two conditions are shown separately. Data from induceddepth adaptation stimulus conditions are shown as open circles, and filled circles for conventional adaptation conditions. Error bars show one
standard error.
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
141
Fig. 5. Individual observer data from the induced-depth test conditions (part 2) are shown in (a–d) along with the mean slope of these data. The
diagonal dash-dot line indicates the theoretical induced depth. Data from induced-depth and conventional adaptation stimulus conditions are shown
as per Figure 4, though this part of the experment was identical in both. Graphs (e–h) show individual aftereffects.
effects were not due to adaptation to measured horizontal retinal disparities. One explanation is that depth
aftereffects are mediated by adaptation at the level of a
3D shape sensitive mechanism. A further logically possible explanation of these data is that the aftereffects in
the conventional adaptation condition were mediated by
adaptation to horizontal disparities, and those in the
induced-depth condition were mediated by adaptation
to vertical disparities. We consider this possibility in the
general discussion.
The fact that aftereffects were very similar across induced-depth and conventional adaptation stimulus
conditions can be taken as evidence that the method
used for matching the induced-depth adaptation stimulus with a theoretically apparently identical conventional stimulus was valid. We explicitly tested this in
Experiment 3, and found evidence that the conventional
and induced-depth stimuli were perceptually equivalent.
We note that in making their postadaptation AFPP
judgments, observers always found that there was a
horizontal disparity modulation which would produce
an apparently frontal stimulus (there was never any residual depth curvature which could not be nulled). This
indicates that depth aftereffects here had the same absolute magnitude as the depth of the stimulus used to
null it, all the way across the stimulus in a horizontal
direction.
3. Experiment 2
Evidence from Experiment 1 suggests that the apparent 3D shape of a stereoscopic adaptation stimulus is
a determiner of depth aftereffects, rather than the underlying horizontal disparity pattern. If so, we would
expect no systematic depth aftereffects following adaptation to apparently frontal stereoscopic stimuli produced by different combinations of horizontal and
vertical disparity modulations. We tested this hypothesis
in Experiment 2.
3.1. Method
3.1.1. Subjects
Three volunteers participated in Experiment 2; two
males and one female, aged between 24 and 27. All
observers had also participated in Experiment 1. Observer PD was one of the authors.
3.1.2. Apparatus and stimuli
The apparatus was identical to that in Experiment 1.
The test stimuli were identical to the Ôconventional test
stimuliÕ used in parts 1 and 3 of Experiment 1. Adaptation stimuli consisted of five disparity patterns: vertical disparities corresponding to fixation distances of 26.8,
34.3, 45, 65.6 and 140.6 cm. Corresponding theoretical
142
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
apparent depths were )4.55, )2.1, 0, 2.1 and 4.55 cm for
a physically frontal plane at the fixation distance of 45
cm. For each adaptation stimulus, a pattern of horizontal disparities (as in Eq. (1)) was chosen from a set to
create an apparently frontoparallel stimulus. The set of
horizontal disparity modulations was exactly the same
as that used for depth nulling in Experiment 1.
3.1.3. Procedure
All observers performed five experimental sessions
of approximately 45 min. One adaptation condition
was examined in each session. Sessions consisted of three
parts. In part 1, observers performed the AFPP task as
in Experiment 1. After a 1 min. break, part 2 started. In
part 2, observers were presented with one of the five
vertical disparity modulated stimuli, which initially had
a horizontal disparity modulation randomly chosen
from the set. The observerÕs task was to set the stimulus
to appear frontal by choosing a pattern of horizontal
disparities from the set using the method of adjustment.
This was performed in real-time and without a time
constraint. Once the observer responded that they were
satisfied with their setting, part 3 began immediately. In
part three observers adapted to the stimulus that they
had just set to apparent frontoparallel, for 4 min. During this time, observers were free to continue tweaking
the depth of the stimulus should it appear to drift from
frontal. The initial adaptation period was followed by
cycles of test trials and adaptation top-up periods of 30
s, as in Experiment 1. Before performing experimental
sessions, observers were given practice in making AFPP
judgements in real-time, as in part 2. Observers performed only one session per day.
3.2. Results and discussion of Experiment 2
In Fig. 6(a)–(c) we show depth aftereffects
(ZAFPP part3 ZAFPP part1 ) for each observer, and the group
mean in (d). Aftereffects were close to zero. We also
recorded time-stamped responses throughout adapta-
tion periods in part 3, allowing us to examine any drift
in the AFPP setting over time. Observers typically made
no adjustments during part 3, or a few minor adjustments (within 0:7 cm of the initial setting). There was
no evidence of a drift in settings between the start and
end of part 3.
The absence of a systematic aftereffect in Experiment
2, unlike Experiment 1, is consistent with our hypothesis
that aftereffects here should depend on the apparent 3D
shape of the adaptation stimulus and not on the horizontal disparity pattern. The finer scale sampling of our
manipulation in this experiment vs. Experiment 1 did
not reveal any aftereffects which were inconsistent with
our hypothesis. Another possible account of the results
is that horizontal and vertical disparity adaptation occurred in equal magnitudes such that any aftereffects
were equal and opposite, and hence cancelled each other
out. We examine this in the general discussion.
The finding of a consistently positive aftereffect
(around 1 cm) for observer CW, which is not predicted
from adaptation to an apparently frontal object, is an
interesting result. One possible account of this is that
since vergence eye position is used in the frontal plane
constancy process (Helmholtz, 1910; Rogers & Bradshaw, 1995), the aftereffect is a result of vergence adaptation (e.g. Ebenholtz & Wolfson, 1975). Any effect of
vergence adaptation should be constant over all our
conditions (in all experiments) since viewing distance
was constant throughout.
4. Experiment 3
The results of Experiments 1 and 2 suggest that the
apparent shape of the adaptation stimuli, rather than
horizontal disparity, is the important determiner of
depth aftereffects. These experiments required that our
adaptation stimuli had the same apparent depths, but
different patterns of horizontal disparity. This was a
simple matter in Experiment 2, as all adaptation stimuli
Fig. 6. Results of Experiment 2. Depth aftereffects for each observer (a–c), and the group mean (d). Standard error bars are smaller than the plot
symbols.
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
were first set by the observer to appear frontal. However, in Experiment 1, our conventional and induceddepth adaptation stimuli were not equated perceptually
(e.g. by viewing each alternately in a matching task), but
instead were equated indirectly from AFPP judgements
in conventional and induced-depth conditions. Because
the depths of these conventional adaptation stimuli were
determined indirectly, one cannot be certain that the
apparent depth of these stimuli was indeed the same as
the apparent depth of the induced-depth adaptation
stimuli. 4 This possibility was examined in Experiment
3, by comparing indirectly matched depth values obtained as in Experiment 1 with directly matched depth
values obtained in a perceptual matching task.
One way that the apparent depth of the resulting
conventional stimuli could have differed from the induced-depth stimuli is if the vertical disparities our
stimuli influenced the estimate of viewing distance used
to scale horizontal disparities in the computation of
metric depth (see Frisby et al., 1999). This is reasonable,
since the vertical disparity field provides a cue to viewing
distance. The effect of changing the estimate of viewing
distance in metric depth scaling is to change the magnitude of the reconstructed depth from a given value of
horizontal disparity (as in Eq. (5)). Such an effect would
result in depth compression of the induced-depth stimuli
for ds < d and an expansion for ds > d, with no distortion when vertical disparities are veridical ðds ¼ dÞ.
Importantly, however, the AFPP task is not influenced
by metric scaling (it is a relief task in the terminology of
G
arding et al.) so any such distortions would not influence the result of the indirect matching process.
Therefore, induced-depth stimuli would appear compressed/expanded in depth relative to the theoretically
matched conventional stimuli. By using a perceptual
matching task in which conventional stimuli are matched to induced-depth comparison stimuli, settings will
reflect any differential effect of vertical disparity on
metric scaling, whereas indirectly matched depth values
will not. If the two sets of data are the same, (in essence,
if the absolute magnitude of the set depth required to
null the apparent depth at one vertical disparity level is
the same as that needed to match that apparent depth at
another vertical disparity level) we can conclude that
vertical disparites did not influence metric scaling, and
that our indirect matching procedure was valid.
4
In Experiment 1, we could in theory have used a direct shape
matching task to obtain conventional stimuli of the same apparent
depth as our induced-dept stimuli. However, because the target surface
must be seen and memorised before each judgement is made––which is
not so in an AFPP task––this and the comparatively greater subjective
difficulty of the task would result in what we felt would be a prohibitively long session duration, without making detrimental sacrifices to
the design in other regards.
143
4.1. Method
4.1.1. Subjects
Four subjects participated in this experiment; all had
normal or corrected visual acuity, and stereoacuity of
<30 arcsec: KM, PD and CW were also participants of
Experiments 1 and 2. PD and LW were the authors.
4.1.2. Apparatus and stimuli
The apparatus was identical to that used in Experiments 1 and 2. Stimuli comprised both the conventional
test stimuli and induced-depth test stimuli as described
in the method for Experiment 1. As in Experiment 1, the
theoretical apparent depths of the induced-depth stimuli
were )4.55, 0 and 4.55 cm. In each session, a conventional stimulus was generated with depth theoretically
matched to the apparent depth of whichever of these
induced-depth stimulus conditions was tested in a given
session, as in Experiment 1.
4.1.3. Procedure
The experiment was of a 1-factor repeated measures
design. Each of the three induced-depth conditions were
examined in separate sessions. Each session comprised
three parts. In part 1, observers made AFPP judgements
using conventional test stimuli, exactly as in part 1 of
Experiments 1 and 2. In part 2, observers made AFPP
settings in one of the induced-depth conditions, exactly
as in part 2 of Experiment 1. From the AFPP settings in
parts 1 and 2, the program calculated the depth of a
conventional stimulus which in theory is perceptually
equivalent to an induced-depth adapting stimulus (as
described in the method for Experiment 1) with the same
vertical disparity modulation as the stimuli in part 2.
This procedure is identical to Experiment 1. In part 3,
observers performed a shape matching task, the induced-depth stimulus just described was used as the
comparison. This was presented for 2 s, followed by a
conventional depth test stimulus, also for 2 s. Observers
responded according to whether the test stimulus had
more or less depth than the comparison. Depth-matching proceeded using a single staircase method with 28
reversals as for parts 1 and 2. Sessions lasted approximately 30 min.
4.2. Results and discussion of Experiment 3
The estimated indirectly matched conventional stimulus depth values obtained using AFPP settings in parts
1 and 2 were compared to observed matched conventional stimulus depth values obtained in part 3. Data are
shown as scatter plots in Fig. 7, for each individual (a)–
(d) and for the group (e). Clearly, estimated and observed matched depth values are similar, lying close to
the diagonal line indicating perfect correspondence. A
paired sample t-test was performed on the data in each
144
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
Fig. 7. Results of Experiment 3 for individual observers (a–d) and group mean (e). These scatterplots show the observed set depth of conventional
stimuli which were matched to induced-depth stimuli, vs. the estimated matched depth obtained indirectly (see text). Identical corresponding values
lie along the diagonal line. In (e) X and Y standard error bars are smaller than the plot symbols.
of the three induced-depth stimulus conditions. In all
cases, estimated and observed matched depth values
were not statistically different: ()4.55 cm: df ¼ 3,
t ¼ 2:381, n.s.; 0 cm: df ¼ 3, t ¼ 1:307, n.s.; 4.55 cm:
df ¼ 3, t ¼ 0:969, n.s.). These results suggest that
vertical disparities did not influence metric scaling of
depth differently between conditions. Had vertical disparities influenced metric scaling differently for the two
types of stimuli, we would have expected the depth
amplitude settings in part 3 to differ from the estimated
indirectly matched values. This was not found. We
conclude that the method used for matching the apparent depth of induced-depth and conventional stimuli
was valid.
We note that even though depth matches in part 3
were made using a specific set of test stimulus surfaces
(as described in Section 1.1), all observers remarked that
their matches could be made satisfactorily. They did not
feel the need to manipulate the shape of the surfaces
along some other dimension in order to equate them
with the induced-depth comparison. We take these results as evidence that the conventional and induceddepth stimuli are metameric (see Backus, 2002).
5. General discussion
The results of Experiments 1 and 2 suggest that depth
aftereffects are dependent on the apparent 3D shape of
an adaptation stimulus, and not due to adaptation of a
horizontal disparity sensitive mechanism. Our results
are consistent with those of Domini et al. (2001), who
reached a similar conclusion. However, it cannot be
determined from either study at what level of 3D shape
processing adaptation may occur; these aftereffects may
have resulted from adaptation of e.g. a metric depth
representation derived from multiple cues, and/or at an
earlier level, e.g. a relief (relative depth) representation
(see G
arding et al., 1995). Indeed, adaptive changes in
the way in which depth cues are combined (see Poom &
B€
orjesson, 1999) could be a factor. Whilst we cannot say
whether the adaptation involved metric and/or nonmetric 3D shape representations, given the finding that
depth aftereffects can transfer across depth cues (Balch
et al., 1977; Bradshaw & Rogers, 1996; Poom & B€
orjesson, 1999), it seems likely that adaptation involved a 3D
shape sensitive mechanism at or beyond the point of
combination of stereo with other depth cues.
One alternative account of our results of Experiments
1 and 2 is that adaptation could occur in both horizontal
and vertical disparity sensitive mechanisms. That is, just
as adaptation to one value of horizontal disparity might
cause errors in subsequent measurement of another value of horizontal disparity (e.g. Howard & Rogers,
1995), the same principle might also apply to vertical
disparities. In this way, in Experiment 1 the conventional aftereffects could be mediated by horizontal
disparity adaptation (and not by vertical disparity
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
adaptation, since adaptation and test stimuli had the
same vertical disparities), and the induced-depth aftereffects could be mediated by vertical disparity adaptation (since adaptation and test stimuli had different
vertical disparities). This account requires the assumption that horizontal and vertical disparity sensitive
mechanisms are both adaptable and in exactly the same
way, in order that the idiosyncratic features of the
conventional aftereffect data appear consistently in the
induced-depth aftereffect data, as we have found in
Experiment 1. This consistency is evident from the data
in Fig. 5(e)–(h) which shows different patterns of aftereffects for each observer, but the same patterns across
conventional and induced-depth adaptation stimulus
conditions. Alternatively however, the perceptual
equivalence of the stimuli (as confirmed by Experiment
3) across conventional and induced-depth conditions
(noting that they are derived from very different disparity patterns) could explain the remarkable similarity
of the idiosyncratic features in our aftereffect data across
the conventional and induced-depth conditions. This
adaptation to apparent depth account, rather than the
adaptation to disparity account, is more consistent with
Domini et al.Õs findings since their stimuli produced aftereffects which varied with apparent depth whilst horizontal disparities were held constant.
A recent study by Berends and Erkelens (2001b) extensively examined adaptation to various patterns of
horizontal and vertical disparities where the appearance
of the stimulus was frontal, in a manner similar to that
in Experiment 2. In one experiment, they examined
adaptation to stimuli with a similar vertical disparity
manipulation to that used here (their transformation
was a quadratic function of both horizontal and vertical
eccentricity) and like in Experiment 2, apparent depth
was nulled by horizontal disparities. Unlike Experiment
2, they found that adaptation to these stimuli did produce systematic aftereffects of depth curvature (this
difference may be due to an important difference in the
test stimuli used, as we shall explain). Berends and
Erkelens concluded that these systematic aftereffects
were not likely due to adaptation of a mechanism sensitive to retinal disparities, since their adaptation phases
involved roving rather than steady fixation. They suggested two alternative explanations of their results: (1)
that adaptation occurred in a mechanism sensitive to
headcentric disparity (see Erkelens & van Ee, 1998), or
(2) that adaptation occurred in the form of recalibration
of the relationship between vertical disparities and eye
position as a result of conflict between the two signals.
The former account seems problematic to us since depth
aftereffects are retinotopic (K€
ohler & Emery, 1947),
suggesting that the relevant adapting mechanism is not
using headcentric coordinates. We suggest that vertical
disparity-driven recalibration of the eye position signals
used for computation of surface slant/curvature (if not
145
also egocentric direction and distance) could account for
the results of Berends and ErkelensÕ study and those of
Experiment 2. We shall refer to the above as the Ôeye
position recalibration hypothesisÕ for short.
An important difference between the two studies is
that Berends and ErkelensÕ test stimuli did not contain
useful vertical disparity information, whereas ours did.
Based on the findings of Backus, Banks, van Ee, and
Crowell (1999) and Rogers and Bradshaw (1995), we
would expect that where useful vertical disparity information is not available, as in Berends and ErkelensÕ test
stimulus, horizontal disparities are interpreted largely
using extraretinal eye position signals, so a relatively
large aftereffect due to the recalibration of these eye
position signals is expected when using such a stimulus.
However, where useful vertical disparity information is
available, as in the test stimulus used in Experiment 2,
horizontal disparities are interpreted largely using vertical disparities. Since the contribution of eye position in
interpreting horizontal disparities is relatively small, the
aftereffect should be relatively small. Indeed the data
showed no measurable systematic aftereffect, consistent
with a low weighting for eye position signals. In this
way, the eye position recalibration hypothesis could
account for both findings, however, it is perhaps surprising that no systematic aftereffect was measured since
the larger estimates of eye position weighting in the literature, for similar stimuli, would be expected to produce some effect. Backus et al. (1999) obtained estimates
of individualsÕ eye position weightings between around
5% and 40% using a slant estimation task with stimuli of
comparable size to those used here. There are several
possible explanations. It may be that our stimulus configuration did not elicit a degree of eye position recalibration that was measurable with our test stimulus; that
our subjects had relatively low eye position weightings
or that eye position signals are purposefully downweighted post-recalibration.
Whilst the eye position recalibration hypothesis
could explain of the results of Experiment 2, a disparity adaptation account is also logically possible, in
which the adaptation stimuli used, and those used
by Berends and Erkelens, resulted in adaptation to
both horizontal and vertical disparities in an identical
manner, such that the two effects cancel each other out.
Note that the same account of identical adaptation to
both horizontal and vertical disparities was logically
possible for Experiment 1 as discussed earlier in this
section. Further empirical work is needed to test the
two accounts.
An important issue relating to the present study is the
role of vertical disparities in metric scaling of depth.
Conclusions on this matter in the literature are mixed
(Frisby et al., 1999; Rogers & Bradshaw, 1993). The
issue arose here because of our method used to match
the apparent depth of induced-depth adaptation stimuli
146
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
with conventional adaptation stimuli in Experiment 1.
This required the assumption that vertical disparities in
our stimuli do not greatly affect metric scaling. If metric
scaling were affected, our induced-depth stimuli would
have appeared compressed/expanded in depth relative
to the theoretically matched conventional stimuli, and
hence we would expect aftereffect magnitude to vary
accordingly. Since induced-depth and conventional aftereffects were almost identical, we concluded that any
effect of vertical disparity on metric scaling was at most
negligible. More direct evidence in support of the perceptual equivalence of our conventional and induceddepth stimuli was found in Experiment 3. Here we found
that the depth of conventional stimuli matched perceptually to induced-depth stimuli were not significantly
different from indirectly obtained theoretical matched
values. Since perceptual matches will reflect an effect of
vertical disparity on metric scaling, but indirect matches
do not, the finding of no difference between the two
indicated no effect here. Evidence in support of the
perceptual equivalence of stereoscopic stimuli consisting
of different patterns of horizontal and vertical disparities
has also been demonstrated by Backus (2002), in the
case of horizontal and vertical uniocular image magnifications.
6. Conclusions
In summary, we found that stereoscopic stimuli
consisting of different combinations of horizontal and
vertical disparity modulations can give rise to the same
apparent curved surfaces in depth, hence, the vertical
disparities of these stimuli were not used as a cue to
distance for the purpose of metric scaling. Adaptation to
various perceptually equivalent curved objects in depth
produced by different disparity patterns gave rise to
identical depth aftereffects. This finding cannot be
accounted for by adaptation of a horizontal disparity
sensitive mechanism alone; instead we argue that the
relevant adapting mechanism is beyond the level of
disparity measurement, and involves representations of
3D shape derived from multiple cues. The results of the
present study taken together with those of Domini et al.
(2001) so far fail to reveal a circumstance in which a
depth aftereffect is produced that can only be attributed
to horizontal disparity adaptation.
Acknowledgements
The authors would like to thank our participants in
this study. We also wish to thank Ben Backus and two
anonymous reviewers for their helpful comments in
preparation of this paper.
Appendix A. General method for stimulus generation
Given the analysis in Section 1.1, the stimuli used in
our AFPP and curvature matching tasks comprised a set
of stereo images which varied along two dimensions: (1)
vertical disparity, i.e. induced-depth levels (independent
variable) and (2) horizontal disparity levels (dependent
variable) used to null apparent depth. Both were obtained by choosing values for a single parameter, simulated vergence distance ds , though of course horizontal
disparity levels were more finely sampled. Where ds
was the same for both, that stimulus corresponded to a
physically frontal surface at that simulated vergence
distance.
The method we set out here ensured that all the
simulated points within each stimulus had the same 2D
configuration across all stimuli in order that there were
no unwanted 2D cues. Generation proceeded as follows:
An initial set of points was generated in the fixation
plane at the veridical screen distance used (45 cm), and
rays from these were projected onto the left and right
retinae. The retinal coordinates of these initial points
were recorded as horizontal and vertical angles in
each eye (the same conventions as used in Mayhew &
Longuet-HigginsÕ (1982) equations). The horizontal and
vertical retinal directions of these rays averaged across
the two eyes gave the mean binocular directions of the
simulated points in space, which remained constant
across all of our stimuli, thus keeping their 2D configuration constant. This was achieved as follows: for each
value of ds , the eyes were rotated to that vergence distance, then a new set of simulated points were generated
in the new fixation plane, whose mean binocular directions were the same as the initial ones. The horizontal
and vertical retinal directions in each eye arising from
viewing this physically frontal plane at ds were then
recorded. The result was a set of horizontal and vertical
retinal directions for left and right eyes viewing frontal
planes over a set of viewing distances. Finally, to produce our stimulus images at the fixation distance of 45
cm, the eyes were rotated back to fixation on the screen
at 45 cm, and the recorded coordinates used to project
rays back on to the screen. Left and right image dots
were plotted at these locations. Horizontal and vertical
disparity modulations in these images could be changed
independently using horizontal and vertical recorded
retinal coordinates obtained from different values of ds .
Note that our independent horizontal and vertical
disparity manipulations in retinal coordinates produced
independent horizontal and vertical changes to the
stimulus images. This is demonstrated by taking any
point in space and moving it along a line passing
through the cyclopean eye (for simplicityÕs sake––this
line approximates the true path of constant mean binocular direction) whilst keeping vergence constant; the
horizontal directions of the projected points in each
P.A. Duke, L.M. Wilcox / Vision Research 43 (2003) 135–147
eyeÕs image plane change oppositely, but their vertical
directions remain unchanged. At the same time, in a
frontal plane (i.e. the screen) the projected points for the
left and right eye shift oppositely and horizontally
(along a raster line), but do not shift vertically. In this
way it is evident that our horizontal disparity manipulations produced only a horizontal displacement of
corresponding left and right eyesÕ images on the screen.
References
Adams, W., Frisby, J. P., Buckley, D., G
arding, J., Hippisley-Cox, S.
D., & Porrill, J. (1996). Pooling of vertical disparities by the human
visual system. Perception, 25, 165–176.
Backus, B. T. (2002). Perceptual metamers in stereoscopic vision. In T.
G. Dietterich, S. Becker, & Z. Ghahramani (Eds.), Advances in
neural information processing systems 14. Cambridge: MIT Press.
Backus, B. T., Banks, M. S., van Ee, R., & Crowell, J. A. (1999).
Horizontal and vertical disparity, eye position, and stereoscopic
slant perception. Vision Research, 39, 1143–1170.
Balch, W., Milewski, A., & Yonas, A. (1977). Mechanisms underlying
the slant aftereffect. Perception and Psychophysics, 21(6), 581–585.
Banks, M. S., & Backus, B. T. (1998). Extra-retinal and perspective
cues cause the small range of the induced effect. Vision Research,
38, 187–194.
Bednar, J. A., & Miikkulainen, R. (1997). A neural network model of
visual tilt aftereffects. In Proceedings of the 19th annual conference
of the cognitive science society (CogSci-97, Stanford, CA) (pp. 37–
42). Hillsdale, NJ: Erlbaum.
Berends, E. M., & Erkelens, C. J. (2001a). Strength of depth effects
induced by three types of vertical disparity. Vision Research, 41,
37–45.
Berends, E. M., & Erkelens, C. J. (2001b). Adaptation to disparity but
not to perceived depth. Vision Research, 41, 883–892.
Bergman, R., & Gibson, J. J. (1959). The negative after-effect of the
perception of a surface slanted in the third dimension. American
Journal of Psychology, 72, 364–374.
Blakemore, C., & Campbell, F. W. (1969). On the existence of
neurones in the human visual system, selectively sensitive to the
orientation and size of retinal images. Journal of Physiology, 203,
237–266.
Blakemore, C., & Julesz, B. (1971). Stereoscopic depth aftereffect
produced without monocular cues. Science, 171, 286–288.
Bradshaw, M. F., & Rogers, B. J. (1996). The interaction of binocular
disparity and motion parallax in the computation of depth. Vision
Research, 36(21), 3457–3468.
147
Coltheart, M. (1971). Visual feature-analyzers and after-effects of tilt
and curvature. Psychological Review, 78(2), 114–121.
Domini, F., Adams, W., & Banks, M. S. (2001). 3D after-effects are
due to shape and not disparity adaptation. Vision Research, 41,
2733–2739.
Ebenholtz, S. M., & Wolfson, D. M. (1975). Perceptual aftereffects of
sustained convergence. Perception and Psychophysics, 17(5), 485–
491.
Erkelens, C. J., & van Ee, R. (1998). A computational model of
perception based on headcentric disparity. Vision Research, 38,
2999–3018.
Foley, J. M. (1980). Binocular distance perception. Psychological
Review, 87, 411–434.
Frisby, J. P., Buckley, D., Grant, H., G
arding, J., Horsman, J. M.,
Hippisley-Cox, S. D., & Porrill, J. (1999). An orientation anisotropy in the effects of scaling vertical disparities. Vision Research,
39, 481–492.
G
arding, J., Porrill, J., Mayhew, J. E. W., & Frisby, J. P. (1995).
Stereopsis, vertical disparity and relief transformations. Vision
Research, 35(5), 703–722.
Gibson, J. J., & Radner, M. (1937). Adaptation, after-effect and
contrast in the perception of tilted lines. I. Quantitative studies.
Journal of Experimental Psychology, 20, 453–467.
Helmholtz, H. von (1910). Handbuch der physiologischen Optik (3rd
ed.). Hamburg: Voss (Transl. Southall, J. P. C. (1962). Physiological optics (Vol. 3). New York: Dover).
Howard, I. P., & Rogers, B. J. (1995). Binocular vision and stereopsis.
Oxford: Oxford University Press.
K€
ohler, W., & Emery, D. A. (1947). Figural after-effects in the third
dimension of visual space. American Journal of Psychology, 60(2),
159–201.
Long, N. R., & Over, R. (1973). Stereoscopic depth aftereffects with
random-dot patterns. Vision Research, 13, 1283–1287.
Mayhew, J. E. W., & Longuet-Higgins, H. C. (1982). A computational model of binocular depth perception. Nature, 297, 376–
378.
Mitchell, D. E., & Baker, A. G. (1973). Stereoscopic aftereffects:
evidence for disparity-specific neurones in the human visual system.
Vision Research, 13, 2273–2288.
Poom, L., & B€
orjesson, E. (1999). Perceptual depth synthesis in the
visual system as revealed by selective adaptation. Journal of
Experimental Psychology: Human Perception and Performance,
25(2), 504–517.
Rogers, B. J., & Bradshaw, M. F. (1993). Vertical disparities,
differential perspective and binocular stereopsis. Nature, 361,
253–255.
Rogers, B. J., & Bradshaw, M. F. (1995). Disparity scaling and
the perception of frontoparallel surfaces. Perception, 24, 155–
179.
Download